Google’s TurboQuant Cuts Memory Use 6x, Keeps Nvidia HBM Stable
Google’s TurboQuant reduces memory use in LLM inference by 6x, prompting investors to sell flash and storage chip stocks while HBM for Nvidia accelerators remains stable. Nvidia CEO Jensen Huang joins the President’s Council of Advisors on Science and Technology following David Sacks’ departure as crypto czar.
1. Google’s TurboQuant Drives Memory Market Divergence
Google unveiled TurboQuant, a technique that cuts memory requirements for large language model inference by a factor of six, triggering a sharp selloff in flash and storage-focused chip stocks. Investors differentiated between storage and high-bandwidth memory, with Samsung, SK Hynix flash shares falling while HBM suppliers tied to Nvidia accelerators saw minimal impact.
2. Nvidia CEO Joins Presidential Science Advisory Council
Following the departure of White House Crypto Czar David Sacks, Nvidia CEO Jensen Huang was appointed to the President’s Council of Advisors on Science and Technology. Huang’s inclusion underscores Nvidia’s influence on U.S. AI policy and could enhance the company’s engagement with federal research initiatives.